The Concorde Trap

Sunk Costs and Asymmetric Loss Aversion in Prospect Theory

On November 29, 1962, the British Minister of Aviation Julian Amery and the French Ambassador Geoffroy de Courcel met in London to sign an intergovernmental treaty binding Sud Aviation of Toulouse and the British Aircraft Corporation of Filton to jointly build a Mach 2.04 supersonic commercial airliner. Named Concorde, the delta-wing jet was projected by ministries in London and Paris to cost a combined £150 million to develop and to capture orders for 250 aircraft. Crucially, at the insistence of the French negotiators—who suspected the British Treasury might lose its nerve once bills arrived—Article 6 of the 1962 treaty contained no cancellation or exit clause. If either nation walked away unilaterally, the withdrawing government could be sued in the International Court of Justice at The Hague for the full financial liability borne by its partner.

Within twenty-four months, the metallurgical and thermodynamic costs of cruising at 1,350 miles per hour at 60,000 feet shattered every projection. Because air friction heated the aluminum skin to 127 degrees Celsius during flight—stretching the airframe by ten inches in midair—engineers had to invent creep-resistant Hiduminium RR.58 skins, computerized air-intake ramps for the four Rolls-Royce Olympus 593 turbojets, and tilting droop-nose visors. By October 1964, when Prime Minister Harold Wilson's Labour government took office in London, the projected development cost had more than doubled from £150 million to £330 million.

By 1973, the economic case for Concorde had collapsed. First, testing proved that Concorde's sonic boom shattered windows along its flight path, prompting American and European regulators to ban supersonic flight over land—eliminating 80 percent of its route network. Second, the October 1973 OPEC oil embargo quadrupled the price of aviation kerosene, turning a four-afterburner jet carrying 100 passengers into an operating disaster compared to Boeing's wide-body 747, which carried 400 passengers at a fraction of the fuel burn per seat. In January 1973, Pan American World Airways and Trans World Airlines canceled all their purchase options; within months, fourteen other carriers followed suit, leaving only two state-owned captive buyers: British Airways and Air France.

In classical microeconomic theory, any rational allocator facing Concorde's numbers in 1973 would have halted construction immediately. A fundamental axiom of rational choice states that historical expenditures that cannot be recovered—sunk costs—are irrelevant to present decisions; only incremental future revenues compared against incremental future costs matter. Yet between 1973 and the aircraft's entry into service on January 21, 1976, British and French ministers poured hundreds of millions of additional pounds into Filton and Toulouse, driving the final development bill to £1.3 billion—nearly nine times the 1962 estimate—to build just fourteen commercial aircraft, none of which ever recouped a single pound of capital expenditure.

Why do ministers and executives throw good money after bad when a project is visibly dead? In March 1979 in Econometrica, two Israeli psychologists from the Hebrew University of Jerusalem, Daniel Kahneman and Amos Tversky, published Prospect Theory: An Analysis of Decision under Risk, replacing Daniel Bernoulli's wealth-utility curve with an empirical map of how human nervous systems evaluate gains and losses.

Through controlled choice experiments, Kahneman and Tversky demonstrated that humans evaluate outcomes not as absolute final wealth, but as changes relative to a reference point—usually the status quo or initial price paid. Across that reference point $x = 0$, the subjective value function $v(x)$ obeys a two-part power law with two asymmetries:

When Kahneman and Tversky measured the empirical parameters ($\alpha \approx \beta \approx 0.88$ and $\lambda \approx 2.25$), two structural traps emerged:

1. 1. Loss Aversion ($\lambda \approx 2.25$): The value curve is more than twice as steep for losses as for gains. Losing £100 million inflicts roughly 2.25 times more pain than gaining £100 million produces pleasure. So long as a doomed project remains open, its cost can be framed as an "ongoing investment." The instant an executive cancels the program, the ledger forces them to crystallize an irreversible loss of $v(-x)$.
2. 2. Risk-Seeking in the Domain of Losses ($\beta < 1$): Because $\beta < 1$, the value function is concave for gains (making people cautious when ahead) but convex for losses. Because the first £500 million of loss already inflicts most of the humiliation, losing an additional £200 million adds little extra marginal pain along the flattening curve. When trapped deep in the loss zone, a decision-maker rejects a sure loss of £500 million and gambles on a double-down carrying a 90 percent chance of losing £700 million and a 10 percent chance of breaking even.

Inside Whitehall and the Élysée Palace in 1973, terminating Concorde meant booking an immediate political loss of £800 million, whereas spending another £500 million kept the unbooked illusion alive. When the British government finally wrote off the debt and handed five Concordes to British Airways for £1 each, the biologists Richard Dawkins and Jane Brockmann named the propensity to waste fatal energy defending a ruined investment the "Concorde fallacy."